Picture a room full of neuroscientists in Porto and imagine one of them standing up and saying, in measured but unmistakable terms, that the thing they have all spent their careers mapping might not actually produce the thing they are trying to explain. That is roughly what happened in April 2026, when Christof Koch took the stage at the 15th 'Behind and Beyond the Brain' Symposium, organized by the BIAL Foundation. Koch, one of the world's most prominent neuroscientists, explored the question of whether the brain creates consciousness at all, noting that the mystery reaches far beyond understanding how neurons process information. The talk arrived not as a career footnote but as a considered provocation from someone who had spent decades inside the machinery: Koch works at the Allen Institute for Brain Science and previously taught at MIT and Caltech, and his research includes developing new ways to identify signs of awareness in patients who appear unresponsive. He knows the wiring better than almost anyone, and that is precisely what makes his skepticism so disorienting.
The target of Koch's challenge has a name philosophers have been using since the 1990s: the "hard problem." Localization is where neuroscience now excels; which regions light up as someone perceives an object, experiences pain, retrieves a memory, or settles on a decision can all be observed, and networks tied to awareness have been pinned down. Yet none of it bridges the distance between electrical and chemical activity and the fact that there is something it is like to be you. Materialism took the brunt of Koch's talk; his question is not whether to dismiss it but whether the account is complete. He organized his critique around three zones of uncertainty, with two of the most pointed being near-death experiences and terminal lucidity, those documented, clinically stubborn cases in which patients with devastated brains, people in the late stages of dementia or prolonged unresponsiveness, suddenly exhibit brief episodes of unexpected clarity, regaining coherent speech, recognition, or purposeful behavior, raising profound questions about the brain at the threshold of death. None of this, in Koch's telling, proves that consciousness can persist without a brain; his argument is that such cases merit serious study, since they might reveal holes in what science can presently explain, and those are two distinct claims worth keeping distinct.
Koch's answer to those holes is not religion and not mysticism; it is a shift in metaphysical baseline. He appeals to extraordinary experiences such as near-death experiences to question physicalism, the perspective that consciousness can be fully reduced to material particles and their interactions, and contra such views, Koch defends idealism and panpsychism, both of which he describes as compatible with naturalism and the scientific method. Koch argues for a 'metaphysical shift,' suggesting that consciousness is a property of the universe like mass or charge. The theoretical vehicle he favors is Integrated Information Theory, or IIT, developed by neuroscientist Giulio Tononi. IIT proposes that any system with a sufficiently high level of integrated information has some form of subjective experience, offering a scientific interpretation of panpsychism. Panpsychism suggests that consciousness is not limited to human beings or animals but may instead be a fundamental property found throughout nature in varying degrees. Koch is careful about what this does and does not mean. As he has said in other conversations, he does not believe a stone is ensouled, but the evidence suggests there are many more systems that have consciousness, possibly all animals, all unicellular bacteria, and at some level maybe even individual cells, and we might be surrounded by consciousness everywhere.
The critics are loud and they are credentialed. A 2023 letter signed by 124 scholars labeled IIT unfalsifiable pseudoscience, though defenders disagree; a major Templeton-funded empirical test that same year gave IIT a slight edge over its rival Global Neuronal Workspace Theory, but results were inconclusive. Philosopher Eric Schwitzgebel, whose March 2026 paper on AI and consciousness lays out the landscape with surgical care, notes that calculating Phi for complex modern neural networks is computationally intractable and that the theory leads to counterintuitive conclusions, such as simple grid-like logic gates having high consciousness, and argues we are caught in an epistemic gap. After decades in neuroscience, Koch himself finds some of the metaphysical claims of physicalism naive; the framework has given science extraordinary explanatory and technological power, but it has left a striking lack of consensus about the nature of consciousness. That admission from inside the establishment is, arguably, the most interesting data point of all.
The practical stakes are no longer abstract. Scientists warn that rapid advances in AI and neurotechnology are outpacing our understanding of consciousness, creating serious ethical risks. IIT pulls the question in a direction that makes AI sentience at once more plausible and more complicated. Advocates of IIT suggest that the most common current computer architectures are unlikely to achieve much consciousness and that consciousness will tend to appear in subsystems of a computer rather than at the level of the computer itself. That is a strange but concrete prediction: a large language model might be nearly blank while some compact feedback loop buried within it flickers with minimal experience. Expert disagreement is not marginal but runs through the center of every serious theory on offer, and Schwitzgebel's goal is not to resolve it but to convince the reader that the disagreement is deep, structurally irreducible in the near term, and existentially consequential. Our existing concepts of consciousness, self, and unified experience are calibrated on vertebrate biology and may fail catastrophically in the AI case. Koch's Porto lecture does not close any of these questions; it tears the floorboards up and invites everyone to look at what is underneath.
If consciousness is a feature of reality rather than a product of tissue, then the question of whether an AI system is sentient is not an engineering problem waiting for better benchmarks; it is a physics problem that humanity does not yet know how to solve.